Journal of Chromatography B, 875 (2008) 180–191
Contents lists available at ScienceDirect
Journal of Chromatography B
journal homepage: www.elsevier.com/locate/chromb
On-column epimerization of dihydroartemisinin: An effective analytical
approach to overcome the shortcomings of the International
Pharmacopoeia monograph夽
Walter Cabri a , Alessia Ciogli b , Ilaria D’Acquarica b , Michela Di Mattia a , Bruno Galletti a ,
Francesco Gasparrini b,∗ , Fabrizio Giorgi a , Silvana Lalli a , Marco Pierini b , Patrizia Simone b
a
b
Analytical Development, R&D Department, sigma-tau S.p.A., Via Pontina km 30,400, 00040 Pomezia, Italy
Dipartimento di Chimica e Tecnologie del Farmaco, Sapienza Università di Roma, P. le Aldo Moro 5, 00185 Rome, Italy
a r t i c l e
i n f o
Article history:
Received 11 April 2008
Accepted 20 June 2008
Available online 27 June 2008
Keywords:
Dihydroartemisinin (DHA)
Antimalarials
Epimerization study
Cryo-HPLC
Dynamic HPLC (DHPLC)
Computer simulation
a b s t r a c t
We developed a cryo-HPLC/UV method for the simultaneous determination of artemisinin (1), ␣dihydroartemisinin (2␣), -dihydroartemisinin (2), and a ubiquitous thermal decomposition product
of 2 (designated as diketoaldehyde, 3), starting from the International Pharmacopoeia monograph on
dihydroartemisinin. The method takes for the first time the on-column epimerization process of 2 into
consideration. Chromatographic separation was obtained under reversed-phase conditions on a Symmetry C18 column (3.5 m particle size) with a mobile phase consisting of acetonitrile–water 60:40 (v/v),
delivered at 0.60–1.00 ml/min flow-rates, with ultraviolet detection at low wavelength ( = 210 nm). Low
temperatures (T = 0–10 ◦ C) were selected on the grounds of a diastereoselective dynamic HPLC (DHPLC)
study performed at different temperatures, aimed at identifying the best experimental conditions capable
of minimizing the on-column interconversion process.
© 2008 Elsevier B.V. All rights reserved.
1. Introduction
Artemisinin (1, Fig. 1) is a sesquiterpene lactone endoperoxide
isolated from Artemisia annua L. that Chinese herbalists traditionally use to treat malaria [1]. Since its identification in the 1970s,
artemisinin, as well as semi-synthetic derivatives [2] and synthetic
trioxanes [3], have been used in therapy. Reduction of artemisinin
by sodium borohydride [4] produced dihydroartemisinin (DHA, 2,
Fig. 1), which is also its main metabolite and provides improved
antimalarial potency and a major elimination route [3]. The synthesis of 2 opened pathways for further derivatization at C-10 to
give ether and ester derivatives, largely exploited by the Chinese
[5] with the aim of tuning water and/or oil solubility and improving
bioavailability. The conversion of the lactone carbonyl group at C-10
of 1 into the hydroxyl (hemiacetal) group in 2 yields a new sterically
labile centre in the molecule, which, in turn, provides two lactol
hemiacetal epimers, namely, 2␣ and 2 (Fig. 2A). The ␣-epimer
bears the hydroxyl group in the equatorial position (absolute stere-
夽 This paper is part of the special issue ‘Enantioseparations’, dedicated to W.
Lindner, edited by B. Chankvetadze and E. Francotte.
∗ Corresponding author. Tel.: +39 06 49912776; fax: +39 06 49912780.
E-mail address: francesco.gasparrini@uniroma1.it (F. Gasparrini).
1570-0232/$ – see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.jchromb.2008.06.037
ochemistry at C-10: R), whereas the -epimer possesses an axial
hydroxyl group [6]. Although 2 has a chair-like pyranose ring, such
nomenclature is the reverse of that normally used for designating
the stereochemistry of sugars and glycosides, in which, for example,
␣-d-glucopyranose possesses an axial hydroxyl group [7].
Bulk crystalline 2 is the 2-epimer (see Fig. 2B), as illustrated
by an X-ray crystallographic study on crystals of 2 [8]. Dissolution
of vacuum-dried solid 2 in CDCl3 provides a solution consisting
exclusively of 2, which equilibrates to a 1:1 mixture of 2␣ and
2 within 10 h [8,9]. The rate and extent of interconversion of the
epimers in solution was shown to be dependent on solvent polarity
[9,10].
Numerous HPLC methods were developed for the analysis and
plasma levels monitoring of 2, formerly based on two main detection strategies: reductive electrochemical (EC) [11–17] and UV
detection [18–20] involving pre- or post-column derivatization. The
latter approach lacks specificity in that metabolites of the drug are
also converted, in many instances, to identical UV-absorbing products. On the other hand, HPLC-EC provides excellent specificity and
sensitivity, although it suffers from some inherent difficulties, i.e.,
rigorous deoxygenation of samples and mobile phases, and special
laboratory facilities are needed. To overcome these shortcomings,
evaporative light scattering detection (ELSD) was coupled for the
first time to the HPLC analysis of artemisinin and related analogues
W. Cabri et al. / J. Chromatogr. B 875 (2008) 180–191
181
Fig. 1. Chemical structures of artemisinin (1), dihydroartemisinin (DHA, 2), and a thermal decomposition product of dihydroartemisinin, designated as diketoaldehyde (DKA,
3).
[21]. The high sensitivity and selectivity of mass spectrometry (MS)
opened the way to a large production of analytical methods based
on the HPLC-MS coupling [22–26] for the plasma monitoring of 2,
either based on atmospheric pressure chemical ionization (APCI)
[23,26] or electrospray ionization (ESI) mode [22,24,25]. Radiochromatographic detection [27] was exploited as well in a recent HPLC
study aimed at determining the 2␣/2 ratio in vivo and evaluating
the protein binding of 2. Notwithstanding such large availability in
the literature of robust HPLC methods suitable for the analysis of
2, only a few were designed for the differential quantitation of the
two isomeric forms of 2 [15,16,18,19,26,27]. As a result, efficient and
robust separation of the two interconverting species must be a prerequisite of any analytical method aimed at quantitating the drug
in active ingredient, pharmaceutical formulations, and biological
fluids. Moreover, since previous studies [28,29] on the equilibrium
between 2␣ and 2 showed that interconversion of the two epimers
occurred in a chromatographic time scale, an ideal HPLC analytical
method for 2 should prevent the epimerization phenomena during
the separation process and allow quantification of the two epimers
even in the presence of related substances. For example, a pharmaceutical batch of 2 can contain several impurities arising from
the specific synthetic procedure [4], such as the starting material
itself (i.e., 1), and a thermal decomposition product, designated as
diketoaldehyde (DKA, 3, Fig. 1) [30,31]. In the present paper we
describe the development of an HPLC/UV method for the simultaneous determination of 1, 2␣, 2, and 3, which, for the first time,
takes into consideration the on-column epimerization process of
2. Starting from the International Pharmacopoeia monograph on
dihydroartemisinin [32], we identified some optimal conditions
(such as stationary phase and column temperature) to minimize
on-column epimerization while achieving the best selectivity and
efficiency of separation. In another related paper, we will evaluate the influence of mobile phase composition to both improve the
overall selectivity and minimize the on-column interconversion.
2. Experimental
2.1. Apparatus
Liquid chromatography was performed using a Waters Model
2695 HPLC separation module (Waters, Milford, MA, USA) coupled
with a Waters 996 Photodiode Array Detector. Chromatographic
data were collected and processed using Empower2 software
(Waters).
Variable temperature HPLC was performed by using a thermally insulated container cooled by the expansion of liquid carbon
dioxide (CO2 ). Flow of liquid CO2 and column temperature were
regulated by a solenoid valve, thermocouple, and electric controller.
Temperature variations after thermal equilibration were within
±0.1 ◦ C.
1 H-NMR spectra were recorded at T = 25 ◦ C on a Varian INOVA
500 MHz spectrometer equipped with a triple resonance indirect
probe (TRIAX). Data acquisition, Fourier transformation, and spectra elaboration were performed using the Varian software VNMR,
6.1C.
2.2. Chemicals and reagents
Fig. 2. Chemical structure (A) and polytube model (B) of the interconverting epimers
of dihydroartemisinin: the 2␣-epimer bears the hydroxyl group in the equatorial
position (absolute stereochemistry at C-10: R), whereas the 2-epimer possesses an
axial hydroxyl group. Polytube model of the 2-epimer was obtained by computer
editing of the X-ray data of crystalline 2 [8]; the model for the 2␣-epimer was
derived by molecular mechanics optimization (MMFF force field as implemented in
SPARTAN’04 1,0,0) by inverting the configuration at C-10.
Artemisinin (1), dihydroartemisinin (DHA, 2), and diketoaldehyde (DKA, 3) samples were supplied by sigma-tau S.p.A., Pomezia
(Italy). HPLC-grade acetonitrile, methanol, and water were purchased from Carlo Erba (Italy). HPLC-grade acetonitrile from Merck
(Darmstadt, Germany) was also tested. Deuterium oxide was from
Sigma–Aldrich (St. Louis, MO, USA). The following commercially
available reversed-phase C18 HPLC columns were used: Zorbax
SB-C18 (Agilent Technologies, Santa Clara, CA, USA); Luna C18
(Phenomenex, Inc., Torrance, CA, USA); YMC-Pack ProC18 RS (YMC
Europe GmbH, Dinslaken, Germany); SunFire C18 (Waters); Symmetry C18 (Waters); Gemini C18 (Phenomenex, Inc.); Acclaim 120
C18 (Dionex Corporation, Sunnyvale, CA, USA); Ascentis Express
182
W. Cabri et al. / J. Chromatogr. B 875 (2008) 180–191
Table 1
Physicochemical data of the tested commercial RP-C18 columns
Column
Zorbax SB-C18
Luna C18
YMC-Pack ProC18 RS
SunFire C18
Symmetry C18b
Gemini C18
Acclaim 120 C18
Ascentis Express C18
Chromolith RP-18
a
b
Dimension
(mm × mm)
Particle
size (m)
Pore size
(Å)
Surface area
(m2 g−1 )
End-capped
150 × 4.6
150 × 4.6
150 × 4.6
150 × 4.6
150 × 4.6
150 × 4.6
150 × 4.6
150 × 4.6
100 × 4.6
5.0
5.0
5.0
3.5
3.5
3.0
3.0
2.7
–
80
100
80
96
100
110
120
90
–
180
440
500
331
340
375
300
150
300
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
K0 × 1014 (m2 )
4.41
2.51
2.36
1.31
1.70
1.30
1.56
1.18
5.95
(%)
(mol m−2 )
Hold-up timea
(min)
–
19.00
21.90
16.62
19.67
14.00
18.00
–
18.00
2.00
3.00
–
3.76
3.15
–
3.20
3.50
3.60
1.56
1.70
1.61
1.63
1.56
1.78
1.79
1.45
1.54
Carbon load
Mean value of 3 injections (sample: nitromethane; eluent: acetonitrile; flow-rate: 1.00 ml/min; T = 25 ◦ C; UV detection at 254 nm).
Also checked in different column lengths (100 mm × 4.6 mm and 75 mm × 4.6 mm I.D.)
C18 (Sigma–Aldrich); Chromolith RP-18 (Merck). Details of the
columns are collected in Table 1.
2.3.2), a purified sample of 2, recrystallized from ethanol, was also
employed, and dissolved in mobile phase as described above.
2.3. Chromatographic procedures
2.5.
2.3.1. Room temperature chromatography
Chromatographic separations were obtained under reversedphase conditions with a mobile phase consisting of acetonitrile–water 60:40 (v/v), delivered at a flow-rate of 1.00 ml/min, with
ultraviolet detection at 210 nm.
Columns temperature was set at T = 25 ◦ C. Columns hold-up time
(t0 ) was determined from the elution time of an unretained marker
(nitromethane) using acetonitrile as eluent, at T = 25 ◦ C, flow-rate
1.00 ml/min and UV detection at 254 nm. Hold-up times, obtained
as mean value of 3 injections, are reported in Table 1.
A 1 mg/ml solution of a purified sample of 2, recrystallized from
ethanol, was prepared in mobile phase, and sonicated to dissolve.
A 0.05 ml aliquot of D2 O was added to 0.6 ml of the above solution, and the mixture was allowed to equilibrate at T = 25 ◦ C. The
final solvent composition was acetonitrile–water 55.4:44.6 (v/v).
1 H-NMR spectra were acquired by means of the DPFGSE solvent
suppression sequence [33], in which the selective pulses were convoluted to obtain simultaneous suppression of the non-deuterated
solvents. 64 scans were acquired, with a recycle delay of 3 s, and
three independent spectra were acquired. From integration of the
singlets at ı = 5.55 and 5.40 ppm, corresponding to the protons at
C-12 for 2 and 2␣ epimers respectively (see Fig. 1), the 2␣ isomer
was 78.6 ± 0.3% (H-NMR K␣/ = 3.5). The very same sample was processed under optimized cryo-HPLC conditions (vide infra), giving a
77.8 ± 0.2% value (HPLC K␣/ = 3.5) for the same isomer.
2.3.2. Variable temperature chromatography
Variable temperature HPLC was performed by placing the
columns inside the device described in Section 2.1. Chromatographic conditions were as reported in Section 2.3.1, except for
column temperature.
1 H-NMR
calculation of the thermodynamic ratio (K˛/ˇ )
2.6. Simulation of diastereoselective dynamic chromatograms
2.3.3. Low-temperature chromatography
Low-temperature separations were performed merely on the
Symmetry C18 column, tested in three different column lengths,
i.e., 150, 100, and 75 mm × 4.6 mm I.D. Chromatographic conditions were acetonitrile–water 60:40 (v/v), flow-rate 0.60 ml/min
(0.60–2.0 ml/min for geometry 75 mm × 4.6 mm I.D.), and UV
detection at 210 nm. Columns temperature was set at T = 0, 5, and
10 ◦ C.
2.4. Sample preparation
Approximately 10 mg, accurately weighed, of a bulk sample of
2 were transferred into a 10 ml glass volumetric flask and then
approximately 10 ml of mobile phase was added. The mixture was
sonicated to dissolve, diluted to volume with mobile phase, and
allowed to equilibrate at T = 25 ◦ C (2␣/2 ratio of about 3.2). A separate solution of 1 was prepared as follows: approximately 10 mg,
accurately weighed, of 1 were transferred into a 5 ml glass volumetric flask and then approximately 5 ml of mobile phase was added.
The mixture was sonicated to dissolve and diluted to volume with
mobile phase. The final concentration of the above solutions was
about 1 mg/ml for 2 and 2 mg/ml for 1. A 200 l aliquot of the latter
solution was added to the former, and a 5 l aliquot of the final mixture was injected into the HPLC system. Compound 3 is a thermal
decomposition impurity which is formed after prolonged storage of
2 [31] and thus always present in small amounts upon dissolution
of 2. For the variable temperature HPLC experiments (see Section
Simulation of variable-temperature experimental chromatograms was performed by using the lab-made computer
program Auto DHPLC y2k [34] which implements both stochastic
and theoretical plate models according to mathematical equations
and procedures described within ref. [35a,b], respectively. A quite
comprehensive view of milestone works on dynamic chromatography and its applications is given in ref. [36]. The developed
algorithm may take into account all types of first-order interconversions, i.e. enantiomerizations as well as diastereomerizations
or constitutional isomerizations, (e.g. pseudo first-order tautomerizations). Within non-enantiomeric isomerizations, forward and
backward interconversion occur at different rates in the achiral
mobile phase, where the two isomerizing species are present
in differing amounts. According to the thermodynamic cycle
involved inside a virtual chromatographic theoretical plate for a
generic first-order isomerization process concomitant with the
chromatographic repartition equilibria (see Chart 1), we applied in
the algorithm the following general equation:
m
k−1
k1m
×
k1s
s
k−1
=
kB′
′
kA
′ and k′ are the retention factors of the first (A) and second
where kA
B
m and km are the rate constants for the back(B) eluting species, k−1
1
ward and forward interconversion in mobile phase, respectively,
s are the rate constants for the forward and backward
and k1s and k−1
interconversion in stationary phase, respectively.
W. Cabri et al. / J. Chromatogr. B 875 (2008) 180–191
183
Chart 1.
Program functionality was validated on several first-order isomerizations (both enantiomerization and non-enantiomerization)
by comparing DHPLC results with those obtained by DNMR technique [34a–d] or by classical method [34e]. The algorithm also
implements the chance of taking tailing effects into account. Both
chromatographic and kinetic parameters can be automatically optimized by simplex algorithm until obtaining the best agreement
between experimental and simulated dynamic chromatograms.
In the present paper all simulations were performed employing
the stochastic model and taking tailing effects into consideration.
3. Results and discussion
3.1. Pharmacopoeia guidelines on antimalarial drugs
Initially, we decided to start our investigation by referring to the International Pharmacopoeia guidelines on
Fig. 3. Typical room temperature chromatograms obtained for a standard mixture of compounds 1 and 2 (containing 3 as impurity). Peak 1 corresponds to 3, peak 2 to the
2␣-epimer, peak 3 to the 2-epimer, and peak 4 to 1. Columns: Zorbax SB-C18, Luna C18, YMC-Pack ProC18 RS, SunFire C18, Symmetry C18, Gemini C18, Acclaim 120 C18,
Ascentis Express C18 (150 mm × 4.6 mm I.D.), and Chromolith RP-18 (100 mm × 4.6 mm I.D.); eluent: acetonitrile–water (60:40, v/v); flow-rate: 1.00 ml/min, T = 25 ◦ C; UV
detection at 210 nm.
184
Table 2
Room temperature chromatographic data for the tested commercial RP-C18 columna
W. Cabri et al. / J. Chromatogr. B 875 (2008) 180–191
a
Eluent: acetonitrile–water = 60:40 (v/v); flow-rate: 1.00 ml/min; T = 25 ◦ C; UV detection at 210 nm; t0, ext. = 0.15 min. Peak 1 = 3; peak 2 = 2␣; peak 3 = 2; peak 4 = 1.
k’, retention factor = (tR − t0 )/t0,corr. = (tR − t0 )/(t0 − t0,extra column ).
c
As : Asymmetry factor.
d
˛: selectivity factor.
e
Rs : USP resolution.
f
Not resolved.
b
185
W. Cabri et al. / J. Chromatogr. B 875 (2008) 180–191
antimalarial drugs [32]. In the monograph on Artenimolum or
Artenimol (i.e., dihydroartemisinin) the method currently recommended is an HPLC assay based on the use of a stainless steel
column (100 mm × 4.6 mm I.D.) packed with a reversed-phase
C18 stationary phase (3.0 m particle size). The mobile phase
is acetonitrile–water 60:40 (v/v), delivered at a flow-rate of
0.6 ml/min and the detection system used is an ultraviolet spectrophotometer set at a wavelength of about 216 nm. The aim of
the pharmacopoeial method is to quantitate dihydroartemisinin
(2) in the presence of artemisinin (1) as related substance. The
monograph does not provide for any other related substance. The
final requirement is that “the test is not valid unless the relative
retention of ␣-artenimol compared with artemisinin is about 0.6,
and the resolution between the peaks is not less than 2.0. Measure
the areas of the peak (twin-peak) responses and calculate the
percentage content of C15 H22 O5 (i.e., artemisinin) with reference
to the dried substance”.
Four main items raised when considering the above method: (i)
the column temperature is not specified, (ii) no mention is made of
the interconversion between the two epimers of 2, which indeed
occurs in the chromatographic time scale, (iii) the conditions are
not very selective towards 3, which is a ubiquitous contaminant of
2, and (iv) signal-to-noise ratios in the presence of plateau zones are
always smaller than in normal elution profiles; as a consequence,
quantitation of species eventually eluting in the plateau would be
negatively affected.
To overcome such shortcomings and try to address the
four points, we investigated nine commercial RP-C18 columns
(150 mm × 4.6 mm I.D.) with different morphological and physicochemical properties, such as specific surface areas (ranging from
150 to 500 m2 g−1 ), particle sizes (from 2.7 to 5.0 m), pore sizes
(from 80 to 120 Å), carbon loads (from 2.00 to 3.76 mol m−2 ), and
permeabilities (from 1.18 to 4.41 × 10−14 m2 ). We also included in
the study a monolithic column (100 mm × 4.6 mm I.D.). Details on
the columns are collected in Table 1. On the nine columns we analyzed the standard mixture of compounds 1 and 2 (containing 3
as impurity), prepared as described in Section 2.4, under pharmacopoeial chromatographic conditions, and at different flow-rates.
Since the column temperature was not specified, we performed
the preliminary chromatographic runs by setting the temperature
at 25 ◦ C. In all cases, we found four chromatographic peaks with
an invariant elution order, which were attributed as follows: peak
1 corresponds to 3, peak 2 to 2␣, peak 3 to 2, and peak 4 to 1.
Typical chromatograms obtained on the tested columns are illustrated in Fig. 3. An interference regime (plateau) between the 2␣
and 2 resolved peaks was detected in all cases, and it is diagnostic of an interconversion process active between the 2␣ (peak
2) and 2 (peak 3) epimers during their chromatographic separation. Different peak shape deformations were observed, depending
on the stationary phase considered. Chromatographic data for the
tested commercial RP-C18 columns are presented in Table 2. As it
can be observed, the investigated peaks were resolved under pharmacopoeial chromatographic conditions, all exhibiting a notable
symmetrical shape (As < 1.42). Selectivity factors (˛) between 3 and
2␣ ranged from 1.12 to 1.19, whereas greater values were reached
for the 2 and 1 couple (˛ between 1.24 and 1.35). The highest selectivities were found for epimer couple 2␣ and 2 (˛ between 1.55
and 1.73), the most selective column being the YMC-Pack ProC18 RS
(˛ = 1.73), followed by the Ascentis Express C18 (˛ = 1.72). Finally,
relative retention of 2␣ (peak 2) compared with 1 (peak 4) was in
good agreement with the pharmacopoeial requirements, ranging
from 0.58 to 0.66 for the conventional particle-packed columns,
reaching 0.73 for the monolithic column.
The main drawback of the method is that the presence of the
visible plateau between the two dihydroartemisinin epimers is
completely neglected. In addition, stationary phases may have a
retarding or activating effect on the kinetics of the dynamic process involving stereolabile species during their passage through
the chromatographic column. Therefore, it was necessary to investigate the epimerization process in the presence of the stationary
phase.
3.2. Evaluation of stationary phases
Since the Pharmacopoeia monograph establishes quantification of dihydroartemisinin from the twin-peak area (i.e., the sum
of ␣- and -artenimol against that of artemisinin used as reference), it seemed relevant to evaluate to what extent the secondary
Table 3
Activation free energies and apparent rate constants for the interconversion process of 2, obtained at T = 25 ◦ C by computer simulation and comparison with the corresponding
off-column data
Column
K␣/ a
Interconversion process 2␣ → 2
app
Symmetry C18d
Symmetry C18e
Symmetry C18f
Luna C18
SunFire C18
YMC-Pack ProC18 RS
Gemini C18
Acclaim 120 C18
Zorbax SB-C18
Ascentis Express C18
Chromolith RP-18
3.1
3.1
3.3
3.3
3.9
4.1
4.3
4.8
4.5
4.4
−
#
G
(kcal mol−1 )
app
kv
(10−2 min−1 )
22.1
22.0
22.0
21.9
21.7
21.6
21.5
21.4
21.2
21.2
21.1
2.30
2.82
2.81
3.52
4.40
5.26
6.60
7.90
9.88
10.2
11.9
mob
Off-columng
a
b
c
d
e
f
g
3.2
K␣/
Interconversion process 2␣ → 2
CESP (%)
app
G#
(kcal mol−1 )
app
kv
(10−2 min−1 )
CESPb (%)
Plateau areaa
(%)
% errorc K␣/
−48.1
−36.3
−36.6
−20.5
−0.7
18.7
49.0
78.3
123.0
130.2
168.6
21.6
21.5
21.5
21.3
21.2
21.2
21.0
20.9
20.7
20.7
20.6
5.38
6.69
6.61
8.62
10.1
11.7
15.9
18.5
23.7
24.3
31.4
−62.4
−53.2
−53.8
−39.7
−29.4
−18.2
11.2
29.4
65.7
69.9
119.6
9.2
5.8
5.0
11.2
21.4
22.5
21.6
24.3
27.2
25.5
15.5
−3.7
−4.2
2.1
1.7
22.6
29.1
34.1
50.2
40.2
36.4
30.4
b
mob
G#
21.7
Calculated by peak areas integrated as depicted in Fig. 5.
Catalytic effect of stationary phase: CESP = [(app kv − mob kv )/mob kv ] × 100.
[(K␣/ − mob K␣/ )/mob K␣/ ] × 100.
Column geometry: 150 mm × 4.6 mm I.D.
Column geometry: 100 mm × 4.6 mm I.D.
Column geometry: 75 mm × 4.6 mm I.D.
Off-column data obtained at T = 25 ◦ C.
mob
kv
4.43
mob
G#
21.0
mob
kv
14.3
186
W. Cabri et al. / J. Chromatogr. B 875 (2008) 180–191
equilibrium between 2␣ and 2 could impact the quantitation
of 2. We therefore performed a preliminary thermodynamic and
kinetic study of the 2␣ ⇄ 2 epimerization process under pharmacopoeial conditions. We checked the equivalence of the UV
absorption coefficients for 2␣ and 2 epimers at the operative
wavelength by comparing chromatographic peak area ratios of 2␣
and 2 at 210 nm (HPLC K␣/ = 3.5) with the corresponding ratios calculated by suitable integrated resonance signals selected within
the 1 H-NMR spectra of 2␣ and 2 (H-NMR K␣/ = 3.5) in the same
solvent mixture (see Section 2.5 for details). Off-column epimerization of solid 2 in mobile phase was performed by incubating
the species into a thermostated reactor and monitoring the process by optimized cryo-HPLC conditions (vide infra). Equilibrium
constant of the 2␣ ⇄ 2 process, mob K␣/ , and pseudo first-order
rate constants for the interconversions of 2␣ to 2, mob kv␣– and 2
to 2␣, mob kv–␣ , were determined at T = 25 ◦ C in mobile phase, and
are collected in Table 3. As expected, the half-lives of both 2␣ → 2
and 2 → 2␣ processes, inferred by the related rate constants at
25 ◦ C, matched the time scale of their separation, thus accounting
for the presence of plateau zones between the corresponding chromatographic peaks. To evaluate the influence of the interconversion
process on the chromatographic performance, we performed lineshape analysis (DHPLC simulations) of all the elution profiles in
Fig. 3. As an example, Fig. 4 shows a comparison of experimental
and simulated chromatograms, obtained for the YMC-Pack ProC18
RS, Symmetry C18, and Chromolith RP-18 columns. As previously
reported [36], rate constants of a given equilibrium determined by
app
dynamic chromatography (the apparent rate constants k1 and
app
k−1 related to the forward and backward reaction, respectively)
′ and k′ ) of
are mean values weighted by the retention factors (kA
B
m ) and stationary
the processes occurring in both mobile (k1m and k−1
s
s
phases (k1 and k−1 ), according to Eqs. (1)–(2).
app
k1
app
=
k−1 =
′
kA
1
m
s
k
+
′ 1
′ k1
1 + kA
1 + kA
(1)
kB′
1
km +
ks
1 + kB′ −1 1 + kB′ −1
(2)
′ and k′ are the retention factors of the first (A) and secwhere kA
B
ond (B) eluting species, respectively. If rate constants in mobile
m are available from independent measurephase, i.e., k1m and k−1
s can
ments, the rate constants in stationary phase, i.e., k1s and k−1
be obtained as well by computer simulation. Therefore, within
the input section of the program, we set rate constants in mobile
m ) as equal to those off-column measured for
phase (k1m and k−1
the 2␣ → 2 and 2 → 2␣ processes in the same media (mob kv␣–
m is always consistent with
and mob kv–␣ ), so that their ratio k1m /k−1
the experimentally measured thermodynamic ratio (K␣/ = 3.2 in
acetonitrile–water 60:40, v/v). Table 3 reports the apparent pseudofirst order rate constants for 2␣ → 2 and 2 → 2␣ conversions
(app kv␣– and app kv–␣ , respectively) and the related activation free
#
app G#
energies (app G␣
), calculated by the Eyring Equa– and
–␣
tion. Also inserted in Table 3 are: (i) the percentage amounts of
the interconverted fractions of 2, responsible for the plateau zones
(ratios between plateau zones area and total 2 area obtained by
integrating from peak 2 to peak 3), and (ii) the K␣/ constants calculated by peaks 3 and 2 area ratio, integrated in presence of the
plateau zone (see Fig. 5 for details on the integration mode used).
By comparison of mob kv␣− and mob kv−␣ with the corresponding app kv␣− and app kv−␣ values, we were able to characterize each
column for its ability to promote or inhibit epimerization of 2.
We defined the Catalytic Effect of Stationary Phases (CESP) as
the percentage increasing (positive) or decreasing (negative) of
the epimerization rate constant, with respect to the analogous
Fig. 4. Comparison of experimental (solid line) and simulated (dotted line) chromatograms. Columns: YMC-Pack ProC18 RS (150 mm × 4.6 mm I.D.), Symmetry C18
(150 mm × 4.6 mm I.D.), and Chromolith RP-18 (100 mm × 4.6 mm I.D.); eluent:
acetonitrile–water (60:40, v/v); flow-rate: 1.00 ml/min, T = 25 ◦ C; UV detection at
210 nm. Simulations were performed by use of the lab-made computer program
Auto DHPLC y2k [34].
value measured in mobile phase (see Table 3). On this basis, three
columns (Symmetry C18, Luna C18, and SunFire C18) of the nine
investigated were shown to reduce the 2␣ ⇄ 2 interconversion
rate, whereas the others increased the rate of the dynamic process, with the exception of the YMC-Pack ProC18 RS column, which
indeed did not show any appreciable effect. According to these findings, columns with negative CESP values (see Table 3) provided
chromatograms with smaller plateau area, that is, a smaller error
in quantitation of 2 can be made when using such columns. In
W. Cabri et al. / J. Chromatogr. B 875 (2008) 180–191
187
Fig. 5. (A) Schematic representation of the integration mode used for the calculation
of 2␣ and 2-epimers area. (B) Computer simulated profiles of 2␣ (dotted line), 2
(dashed line), and of the mixture 2␣ + 2 (solid line).
particular, for the Symmetry C18 and Luna C18 columns, such an
error (plateau area ≤ 11%) is about the half of that found for the other
columns (>21%). A further and definitive confirmation of the convenience in the use of the Symmetry C18 and/or Luna C18 columns
for their inhibiting effect on epimerization derived from the more
precision in the K␣/ constant measurement. Such columns led to
acceptable error ranges (<4%), whereas in the other cases the error
raised to 22%, up to 50%. On the basis of the above considerations,
the Symmetry C18 column was selected to perform a more in-depth
investigation.
3.3. Investigation of the selected Symmetry C18 column
The Symmetry C18 column was tested in three different column
geometries, i.e., 150, 100, and 75 mm × 4.6 mm I.D. As easily predictable, the less the column length, the less is the staying time of
2␣ and 2 epimers inside the separation device. Therefore, lower
plateau areas were obtained when reducing the column length (see
Table 3), with comparable selectivity, but with lower efficiency and
resolution (see Table 2). Typical room temperature chromatograms
achieved on the Symmetry C18 columns are depicted in Fig. 6.
The up-to-now findings clearly demonstrated the role of the
stationary phase (i.e., type and column length) in minimizing
quantitation errors of 2. However, we decided to investigate the
influence of column temperature on the epimerization process,
which is always relevant when considering sterically labile compounds.
Fig. 6. Typical room temperature chromatograms obtained on the Symmetry C18
columns family for a standard mixture of compounds 1 and 2 (containing 3 as impurity). Peak 1 corresponds to 3, peak 2 to the 2␣-epimer, peak 3 to the 2-epimer,
and peak 4 to 1. Columns geometry: A: 150 mm × 4.6 mm I.D.; B: 100 mm × 4.6 mm
I.D.; C: 75 mm × 4.6 mm I.D.; eluent: acetonitrile–water (60:40, v/v); flow-rate:
1.00 ml/min, T = 25 ◦ C; UV detection at 210 nm.
3.4. Diastereoselective dynamic HPLC (DHPLC)
Variable temperature separations were performed on the
Symmetry C18 column, in the geometry recommended by the
Pharmacopoeial monograph (i.e., 100 mm × 4.6 mm I.D.). The temperature range was of 40–0 ◦ C. Since plateau zones were easily
188
W. Cabri et al. / J. Chromatogr. B 875 (2008) 180–191
Fig. 7. Variable temperature chromatographic profiles obtained on the Symmetry C18 column (100 mm × 4.6 mm I.D.). Solid line: experimental chromatograms (eluent:
acetonitrile–water 60:40 v/v, flow-rate: 1.00 ml/min, UV detection at 210 nm). Dotted line: computer simulated profiles obtained with the measured free energy activation
barriers for the on-column epimerization process.
visible only at T > 20 ◦ C, simulations of the chromatograms registered from 40 to 25 ◦ C were performed by using the classical
stochastic model, as implemented within the Auto DHPLC y2k
program [34]. Fig. 7 shows the superimposed experimental and
simulated chromatographic profiles with the measured activation
free energies. van’t Hoff analyses of the obtained data were also carried out to evaluate the enthalpic and entropic contributions to the
epimerization barrier, and the related plots are depicted in Fig. 8.
As judged by the obtained results, just a very small contribution
#
#
#
to G␣
– and G–␣ is due to entropy (S values < 10 u.e.). This
suggests that a monomolecular process must be involved in the
rate-determining step, reasonably, the ring opening of either the
protonated or deprotonated ␣- or -hemiacetalic form of 2, generated in a previous reversible step by reaction with an acid or a base,
respectively. Such kinetic pathway would be in agreement to what
already reported [10].
3.5. Cryo-HPLC on the selected Symmetry C18 columns family
Low-temperature separations were performed on the Symmetry C18 column, in the three different column lengths, i.e., 150, 100,
and 75 mm × 4.6 mm I.D., and at T = 0, 5, and 10 ◦ C. Lower plateau
Fig. 8. van’t Hoff plots obtained by DHPLC for the 2␣ ⇄ 2 interconversion. Column: Symmetry C18 column (100 mm × 4.6 mm I.D.).
Table 4
Low-temperature chromatographic data for the Symmetry RP-C18 columnsa
Column
Symmetry C18
a
b
e
f
T (◦ C)
Peak 1
2 vs 1
tR (min)
k’
b
As
c
d
˛
]Peak 2
Rs
e
3 vs 2
b
tR (min)
k’
As
c
˛
d
Peak 3
Rs
e
4 vs 3
b
tR (min)
k’
As
c
˛
d
Peak 4
Rs
e
tR (min)
k’b
As c
150 × 4.6 (t0 = 2.60 min)
0
5
10
5.92
5.81
5.67
1.41
1.37
1.31
1.07
0.99
0.93
1.15
1.14
1.15
1.98
2.05
2.06
6.41
6.28
6.14
1.62
1.57
1.51
1.10
0.99
0.96
1.70
1.70
1.70
8.91
9.35
9.21
9.09
8.89
8.65
2.76
2.67
2.57
1.04
0.92
0.90
1.32
1.31
1.30
5.74
5.85
5.50
11.18
10.88
10.45
3.65
3.52
3.34
1.04
0.93
0.50
100 × 4.6 (t0 = 1.87 min)
0
5
10
4.15
4.07
3.96
1.41
1.36
1.29
1.13
1.30
1.18
1.13
1.13
1.15
1.07
1.17
1.36
4.44
4.36
4.28
1.59
1.54
1.49
1.14
1.14
1.18
1.73
1.73
1.72
6.42
6.51
7.14
6.34
6.20
6.03
2.76
2.67
2.57
1.14
1.14
1.13
1.35
1.35
1.30
4.47
4.41
4.67
7.90
7.72
7.30
3.72
3.61
3.35
1.15
1.14
1.13
75 × 4.6 (t0 = 1.46 min)
0
5
10
3.20
3.13
3.06
1.43
1.38
1.32
n.r.f
n.r.f
n.r.f
1.12
1.12
1.14
<1
<1
1.03
3.40
3.34
3.28
1.60
1.55
1.50
n.r.f
n.r.f
1.12
1.71
1.72
1.72
5.29
5.36
5.48
4.77
4.69
4.58
2.73
2.67
2.58
1.19
1.18
1.15
1.35
1.33
1.31
3.73
3.64
3.53
5.91
5.77
5.56
3.68
3.56
3.39
1.20
1.19
1.16
Eluent: acetonitrile/water = 60/40 (v/v); flow-rate 0.60 ml/min; UV detection at 210 nm; t0, ext. = 0.25 min. Peak 1 = 3; peak 2 = 2␣; peak 3 = 2; peak 4 = 1.
k’, Retention factor = (tR − t0 )/t0,corr. = (tR − t0 )/(t0 − t0,extra column ).
As : Asymmetry factor.
˛: Selectivity factor.
Rs : USP resolution.
Not resolved.
189
Fig. 9. Cryo-HPLC/UV profiles obtained on the Symmetry C18 column
(150 mm × 4.6 mm I.D.) for a standard mixture of compounds 1 and 2 (containing 3 as impurity). Peak 1 corresponds to 3, peak 2 to the 2␣-epimer, peak 3 to
the 2-epimer, and peak 4 to 1. Eluent: acetonitrile–water 60:40 (v/v); flow-rate:
0.60 ml/min, T = 0 ◦ C (A), 5 ◦ C (B), and 10 ◦ C (C); UV detection at 210 nm.
areas were obtained when decreasing from T = 25 to 0 ◦ C (data not
shown), but slightly lower efficiency and resolution were found, as
judged by the chromatographic data presented in Table 4. Resolution factors (Rs ) between 3 and 2␣ slightly decreased from 2.07 at
T = 25 ◦ C (see Table 2) to 1.98–2.06 at T = 0–10 ◦ C (Table 4) on the
W. Cabri et al. / J. Chromatogr. B 875 (2008) 180–191
c
d
Dimension (mm)
190
W. Cabri et al. / J. Chromatogr. B 875 (2008) 180–191
Fig. 10. Chromatogram profiles obtained at T = 0 ◦ C (left) and T = 25 ◦ C (right) for a methanol solution of 2 heated in an oven at 90 ◦ C for 4 h: an unknown impurity is eluting
in the plateau zone. Column: Symmetry C18 column (100 mm × 4.6 mm I.D.); eluent: acetonitrile–water (60:40, v/v); flow-rate: 0.60 ml/min; UV detection at 210 nm.
150 mm Symmetry C18 column, whereas for the 100 mm column
such decreasing is more relevant (from 1.63 to 1.07–1.36, respectively). A baseline resolution between 3 and 2␣ was achieved only
on the 150 mm Symmetry C18 column. A direct comparison of
the chromatograms obtained on such column at the three column
temperatures further supports the convenience in the use of the
150 mm Symmetry C18 column under cryo-HPLC conditions (Fig. 9)
On the basis of the epimerization rate constants extrapolated at
T < 25 ◦ C, we calculated a marginal plateau area (<2%) only close
to T = 0 ◦ C (Fig. 10, left). We are currently testing buffered mobile
phases to given pHs aimed at minimizing on-column epimerization without drastically decreasing column temperature. No
relevant influence of flow-rates on the chromatographic performances was observed when checked on the Symmetry C18 column
(75 mm × 4.6 mm I.D.), at T = 10 ◦ C (data not shown).
A final consideration on the Pharmacopoeian method can be
made in the case of chemical impurities eventually eluting in the
plateau zone (see Fig. 10, right): quantitation of 2 in such cases
could be difficult to obtain and usually overestimated.
technique to yield fast analysis without compromising the efficiency of separation.
Acknowledgements
Financial support from FIRB, Research program: Ricerca
e Sviluppo del Farmaco (CHEM-PROFARMA-NET), grant no.
RBPR05NWWC 003 is acknowledged.
The authors gratefully acknowledge Giovanna Cancelliere
(Sapienza Università di Roma, Italy) for helpful assistance in the
manuscript editing, and Andrea Mazzanti (Università di Bologna,
Italy) for performing NMR spectra of dihydroartemisinin.
References
[1]
[2]
[3]
[4]
[5]
[6]
4. Conclusions
[7]
A systematic investigation was made on the influence of
chromatographic conditions (stationary phase and column temperature) on the simultaneous reversed-phase HPLC determination of
artemisinin (1), ␣-dihydroartemisinin (2␣), -dihydroartemisinin
(2), and a thermal decomposition product of 2 (diketoaldehyde
3), considering for the first time the on-column epimerization
process of 2. Nine commercial RP-C18 columns with different
pore size, surface area, carbon load and permeability were evaluated starting from the International Pharmacopoeia monograph
on dihydroartemisinin, and compared on the basis of their stationary phase effect towards the interconversion between 2␣ and 2.
Computer simulation was employed to gain insight on the Catalytic
Effect of the Stationary Phase (CESP). Activation free energies of the
process, obtained by computer simulation of the experimentally
obtained elution profiles, showed that the Symmetry C18 column
among the nine analyzed has an inhibiting effect on epimerization
(G# = 22.1 for 2␣ → 2 and 21.6 kcal mol−1 for the backward process). The column was selected for the development of a cryo-HPLC
method aimed at minimizing on-column interconversion. On the
basis of the epimerization rate constants extrapolated at T < 25 ◦ C,
we calculated a marginal plateau area (<2%) only close to T = 0 ◦ C
for the 150 mm × 4.6 column geometry. In another related paper,
buffered mobile phases to given pHs will be tested to minimize
on-column epimerization without drastically decreasing column
temperature. UPLC will be investigated as well as an alternative
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
D.L. Klayman, Science 228 (1985) 1049.
P.M. O’Neill, Expert Opin. Invest. Drugs 14 (2005) 1117.
P.M. O’Neill, G.H. Posner, J. Med. Chem. 47 (2004) 2945.
L. Ying, P. Yu, Y. Chen, L. Li, Y. Gai, D. Wang, Y. Zheng, Chin. Sci. Bull. 24 (1979)
667.
China Cooperative Research Group on qinghaosu, its derivatives as antimalarials, J. Tradit. Chin. Med. 2 (1982) 9.
A.K. Pathak, D.C. Jain, R.P. Sharma, Indian J. Chem., Sect. B: Org. Chem. Incl. Med.
Chem. 34 (1995) 992.
R.K. Haynes, H.W. Chan, M.K. Cheung, S.T. Chung, W.L. Lam, H.W. Tsang, A.
Voerste, I.D. Williams, Eur. J. Org. Chem. (2003) 2098.
X. Luo, H.J.C. Yeh, A. Brossi, J.L. Flipper-Anderson, R. Gillardi, Helv. Chim. Acta
67 (1984) 1515.
R.K. Haynes, H.W. Chan, M.K. Cheung, W.L. Lam, M.K. Soo, H.W. Tsang, A. Voerste,
I.D. Williams, Eur. J. Org. Chem. (2002) 113.
R.K. Haynes, Curr. Med. Chem. 6 (2006) 509.
Z.M. Zhou, J.C. Anders, H. Chung, A.D. Theoharides, J. Chromatogr. 414 (1987)
77.
V. Melendez, J.O. Peggins, T.G. Brewer, A.D. Theoharides, J. Pharm. Sci. 80 (1991)
132.
V. Navaratnam, S.M. Mansor, L.K. Chin, M.N. Mordi, M. Asokan, N.K. Nair, J.
Chromatogr. B 669 (1995) 289.
J. Karbwang, K. Na-Bangchang, P. Molunto, V. Banmairuroi, K. Congpuong, J.
Chromatogr. B 690 (1997) 259.
N. Sandrenan, A. Sioufi, J. Godbillon, C. Netterb, M. Danker, C. van Valkenburg,
J. Chromatogr. B 691 (1997) 145.
V. Navaratnam, M.N. Mordi, S.M. Mansor, J. Chromatogr. B 692 (1997) 157.
C.-S. Lai, N.K. Nair, S.M. Mansor, P.L. Olliaro, V. Navaratnam, J. Chromatogr. B 857
(2007) 308.
K.T. Batty, T.M.E. Davies, L.T. Thu, T.Q. Binh, T.K. Anh, K.F. Ilett, J. Chromatogr. B
677 (1996) 345.
K.T. Batty, K.F. Ilett, T.M.E. Davis, J. Pharm. Pharmacol. 48 (1996) 22.
B.M. Kotecka, K.H. Rieckmann, T.M.E. Davis, K.T. Batty, K.F. Ilett, Acta Trop. 87
(2003) 371.
B.A. Avery, K.K. Venkatesh, M.A. Avery, J. Chromatogr. B 730 (1999) 71.
D. Ortelli, S. Rudaz, E. Cognard, J.-L. Veuthey, Chromatographia 52 (2000) 445.
C. Souppart, N. Gauducheau, N. Sandrenan, F. Richard, J. Chromatogr. B 774
(2002) 195.
S. Sabarinath, M. Rajanikanth, K.P. Madhusudanan, R.C. Gupta, J. Mass Spectrom.
38 (2003) 732.
M. Rajanikanth, K.P. Madhusudanan, R.C. Gupta, Biomed. Chromatogr. 17 (2003)
440.
H. Naik, D.J. Murry, L.E. Kirsch, L. Fleckenstein, J. Chromatogr. B 816 (2005) 233.
K.T. Batty, K.F. Ilett, T.M.E. Davis, Br. J. Clin. Pharmacol. 57 (2004) 529.
W. Cabri et al. / J. Chromatogr. B 875 (2008) 180–191
[28]
[29]
[30]
[31]
Z. Shishan, Chromatographia 22 (1986) 77.
N. Liu, L. Yang, Z. Zhang, Chin. J. Pharm. Anal. 22 (2002) 303.
A.J. Lin, A.D. Theoharides, D.L. Klayman, Tetrahedron 42 (1986) 2181.
L. Dhooghe, S. Van Miert, H. Jansen, A. Vlietinck, L. Pieters, Pharmazie 62 (2007)
12.
[32] International Pharmacopoeia, fourth ed., Monographs for Antimalarial Drugs,
WHO Press, Geneva, 2006, pp. 215–218.
[33] (a) K. Stott, J. Keeler, Q.N. Van, A.J. Shaka, J. Magn. Reson. 125 (1997) 302;
(b) T. Parella, P. Adell, F. Sánchez-Ferrando, A. Virgili, Magn. Reson. Chem. 36
(1998) 245.
[34] (a) F. Gasparrini, L. Lunazzi, A. Mazzanti, M. Pierini, K.M. Pieyrusiewicz, C. Villani, J. Am. Chem. Soc. 122 (2000) 4776;
(b) C. Dell’Erba, F. Gasparrini, S. Grilli, L. Lunazzi, A. Mazzanti, M. Novi, M. Pierini,
C. Tavani, C. Villani, J. Org. Chem. 67 (2002) 1663;
(c) A. Dalla Cort, F. Gasparrini, L. Lunazzi, L. Mandolini, A. Mazzanti, C. Pasquini,
M. Pierini, R. Rompietti, L. Schiaffino, J. Org. Chem. 70 (2005) 8877;
(d) F. Gasparrini, S. Grilli, R. Leardini, L. Lunazzi, A. Mazzanti, D. Nanni, M. Pierini,
M. Pinamonti, J. Org. Chem. 67 (2002) 3089;
191
(e) R. Cirilli, R. Ferretti, F. La Torre, D. Secci, A. Bolasco, S. Carradori, M. Pierini,
J. Chromatogr. A 1172 (2007) 160.
[35] (a) M. Jung, QCPE Bull. 12 (1992) 52;
(b) O. Trapp, V. Schurig, Comput. Chem. 25 (2001) 187.
[36] (a) J.C. Giddings, J. Chromatogr. 3 (1960) 443;
(b) R. Kramer, J. Chromatogr. 107 (1975) 241;
(c) V. Schurig, W. Burkle, J. Am. Chem. Soc. 104 (1982) 7573;
(d) W. Burkle, H. Karfunkel, V. Schurig, J. Chromatogr. 288 (1984) 1;
(e) J. Veciana, M.I. Crespo, Angew. Chem. Int. Ed. Engl. 30 (1991) 74;
(f) O. Trapp, G. Schoetz, V. Schurig, Chirality 13 (2001) 403;
(g) O. Trapp, Anal. Chem. 78 (2006) 189;
(h) J. Oxelbark, S. Allenmark, J. Chem. Soc. Perkin Trans. 28 (1999) 1587;
(i) C. Wolf, Chem. Soc. Rev. 34 (2005) 595;
(j) K. Cabrera, M. Jung, M. Fluck, V. Schurig, J. Chromatogr. A 731 (1996) 315;
(k) R. Kiesswetter, F. Brandl, N. Kastner-Pustet, A. Mannschreck, Chirality 15
(2003) S40.